The CAR construct is critical for activating CAR-transduced cells. CARs used in CAR-NK cells are generally similar to those used in CAR-T cells. A CAR always comprises four components: an extracellular binding domain, a hinge region, a transmembrane domain, and one or more intracellular signaling domains ( Figure 1 ). The extracellular binding domain confers specificity to CAR-modified effector cells by means of targeting TAAs. The hinge region connects the extracellular binding domain with the transmembrane domain. The intracellular signaling domains, which determine the strength of the activation signal and affect the killing activity, are of different compositions in various generations of CARs. The first generation of CARs typically contained only the CD3ζ activation signaling domain, while the second and third generations of CARs combined one or two additional costimulatory molecules (e.g., CD28, ICOS, 4-1BB, CD27, OX40, and CD40). Among these molecules, CD28 and 4-1BB are the most commonly used (33, 34). The fourth generation of CARs, designed by Chmielewski and colleagues, endow the effector immune cells with two transgenic products, the CAR and the transgenic payload. The fourth generation of CARs is focused on removing the current limitations of CAR-based cellular therapy and thus further improving the functions of effector cells (35).

In addition to the common CARs that are applicable for both CAR-T cells and CAR-NK cells, researchers have exploited other molecules that could serve as the activation signaling domain and be more suitable for NK cells. CD244(2B4), a member of the signaling lymphocyte activation molecule (SLAM) family, could also be used as a costimulatory molecule. Increased capacity of signal enhancement and augmented natural cytotoxicity against tumor cells are induced by the upregulation of 2B4 in NK cells (36). DNAX-activation protein (DAP) 12, expressed on NK cells, participates in signal transduction involving NK activating receptors natural-killer group 2 (NKG2) member C (NKG2C) and NKp44; DAP10 also participates in signal transduction involving NKG2D (37, 38). Therefore, DAP12 and DAP10 can transduce intracellular signals in CAR-NK cells. Moreover, NK cells engineered with DAP12-based CARs performed better than NK cells engineered with CD3ζ-based CARs (37). Cumulative evidence demonstrates that NKG2D ligands are overexpressed in several hematological malignancies. Therefore, the CAR construct, NKG2D-DAP10-CD3ζ, which targets the NKG2D ligands, is of great potential in blood cancers. The safety and enhanced killing activity of NKG2D-DAP12-CD3ζ CAR-NK cells for colorectal cancer have been proven in a clinical trial (39).

Various sources of allogeneic NK cells are available for producing CAR-NK cells ( Figure 2 ). Current sources of clinical-grade NK cells include the NK92 cell line, peripheral blood mononuclear cells (PBMCs), umbilical cord blood (UCB), and induced pluripotent stem cells (iPSCs). Human embryonic stem cells (hESCs) and CD34+ hematopoietic progenitor cells (HPCs), also serve as sources of NK cells (40, 41).

The NK92 cell line is commonly used in adoptive immunotherapy. NK92 cells can be easily and reproducibly expanded from a good manufacturing practice (GMP)-compliant cryopreserved master cell bank (42). There are no killer immunoglobulin-like receptors (KIRs) or CD16 expressed on NK92 cells; thus, they cannot mediate ADCC (43). Although NK92 cells exhibit a poorly cytotoxic immunophenotype (CD56^brightCD16^low/-), the NK92 cell line is a highly cytotoxic cell line that expresses high levels of cytolytic pathway molecules (e.g., perforin and granzyme B) (44). Irradiation is necessary before infusion because of their chromosomal abnormalities and risk of malignant transformation (45). A novel type of CAR-NK cell line, CAR-KHYG-1, has shown antitumor effects against glioblastoma cells, suggesting that KHYG-1 may be a new option for CAR-NK cell therapy (46).

Approximately 90% of NK cells isolated from PBMCs are CD56^dimCD16^bright cells, which typically represent a mature population with increased cytotoxicity and reduced proliferative capacity (47). There is no need for irradiation before infusion of PBMC-derived CAR-NK cells because of their lower risk. A great number of NK cells for GMP-grade clinical application can be obtained after isolation, purification, stimulation, and expansion of PBMCs derived from healthy donors or directly from patients themselves (48). However, similar to what happens with autologous T cells, obtaining autologous NK cells may be difficult in patients who have been heavily pre-treated with chemotherapy. Thus, there is a need to discover additional sources of allogeneic NK cells (rather than cells only from healthy donors) as platforms for CAR engineering. Furthermore, the purification is a vital procedure when obtaining allogeneic PBMC-derived NK cells, because residual B cells may cause lymphoma (due to reactivation of Epstein–Barr virus) and residual T cells may cause GVHD (49).

NK cells comprise approximately 30% of lymphocytes in UCB and 10% in the corresponding PB (50, 51). The capacities for cryopreservation and easy collection make UCB-derived NK cells appropriate for off-the-shelf applications (52). There are more than 600,000 UCB units stored in UCB banks, which allows the selection of donors with specific HLA types (53). There are few T cells in UCB and most are immature, thus minimizing the risk of GVHD (54, 55). A large proportion of UCB NK cells exhibit a naïve phenotype with decreased expression of adhesion molecules, CD16, KIRs, perforin, and granzyme B, as well as increased expression of inhibitory molecules (e.g., NKG2A), thus leading to lower cytotoxicity against tumor cells compared with their PB counterparts (52).

iPSCs, derived from non-hematopoietic cells, are another important stem-cell source of allogeneic NK cells. UCB and PBMCs generally cannot provide a homogeneous NK cell population and function in multiple-dosing strategies because it is unpractical to obtain cells from a single renewable source. However, only one single CAR-engineered iPSC cell is sufficient for differentiation into a large number of highly homogeneous CAR-NK cells for clinical use; therefore, repeated doses could be administered to patients to overcome the short NK cell lifespan (19, 52, 56, 57). Similar to NK92 cells and UCB-NK cells, iPSC-derived NK (iNK) cells exhibit an immature phenotype, with low expression levels of KIRs and CD16, as well as higher expression level of NKG2A. Genetically engineering iNK cells to express non-cleavable CD16 Fc receptor to mediate the ADCC effect may be a reasonable solution to this limitation (58, 59). FT596 and FT576, cell products for B-cell malignancies and MM, respectively, are investigational and off-the-shelf CAR-NK cell products derived from a human clonal master iPSC line engineered with anti-tumor abilities. These functional cell products can be manufactured on a large scale to support multi-dose therapeutic strategies and on-demand dose availability (60–62).

Currently, low transduction efficiency caused by the lack of efficient gene transferring approaches is a major barrier. The methods used for transduction of CARs into T cells are also applicable in NK cells. Approaches include viral transduction (retrovirus-based and lentivirus-based) and transfection (electroporation, lipofection, and in combination with transposon systems) (63). Viral-based transduction approaches enable stable integration into the genomes of CAR-NK cells. Retroviral vector transduction exhibits high efficacy (43–93% in primary NK cells), while the insertional mutagenesis and deleterious impact represent major restrictions of this approach in clinical applications (64). Nevertheless, lentivirus-based transduction is safer, while its transduction efficiency (8–16% in PBMC-derived NK cells) must be further improved (65). RNA transfection methods are cost-effective approaches with higher gene transfer efficiency; however, the CAR constructs expression is transient through this method, approximately 3–5 days. On the one hand, the narrow therapeutic time window is a limitation; on the other hand, the incidence of CAR-associated side effects, such as on-target off-tumor effects, may be reduced due to its transient nature (39, 66, 67). Combining transfection methods with DNA integration techniques through transposon systems, such as PiggyBac (PB) and sleeping beauty (SB), has been developed as an attractive approach to generate safer and more stable transgene-expressing cells (68, 69). The SB transposon vector has provided an efficient and economic method for gene transfer, while its applicability in CAR-NK cells remains untested (70).

B-cell originated lymphoma and leukemia which belong to hematological B-cell malignancies, are characterized by the expression of one or more common B-cell antigens (71). CAR-T-based cellular therapy is thriving and promising in the management of B-cell malignancies. Frequently used target antigens in CAR-T cell therapy for B-cell lymphoma and leukemia (e.g., CD19, CD20, and CD22) are also applicable in CAR-NK cells (72, 73). Multiple preclinical and clinical trials have been conducted to test the efficacies of CAR-NK cells for B-cell malignancies ( Table 1 ).

Engineering NK92 cells with an anti-CD19 CAR construct that incorporates a CD3ζ signaling domain either alone or with additional costimulatory molecules (e.g., CD28, 4-1BB, or 2B4) could achieve significant cytotoxicity of NK cells towards B-cell leukemic cell lines and NK-resistant primary leukemia cells in vitro (64, 75–77). These studies have shown that the integration of either 2B4 or CD28 could further enhance all aspects of the CD19 CAR-NK cell activation response to B-cell leukemia cells (76, 77). Moreover, concerning the potential risk of malignant transformation through transfusion of NK92 cells, Liu et al. provided evidence in mouse models that irradiation before infusion of CD19 CAR-NK92 cells could avoid the potential risk of secondary NK lymphoma without the reduction of killing capacity against CD19+ malignancies, compared with unirradiated cells (78). The therapeutic potential and effectiveness of CD20 CAR-NK cells for CD20+ aggressive B-cell non-Hodgkin’s lymphoma (NHL) have also been confirmed in mouse models by Chu et al. (79, 80). Another group designed dual CD19/CD22 CAR-NK cells, which can kill both CD19^KO- and CD22^KO-RS4;11 cells (an ALL cell line) in vitro (81).

Clinical trials registered on http://www.clinicaltrials.gov involving CAR-NK cell therapy in hematological malignancies are summarized in Table 1 . CD19 is a popular target in both CAR-T and CAR-NK cell therapy methods; many clinical trials are focused on B-cell malignancies. Thus far, only one Phase I/II large-scale clinical trial has been published (31). Eleven heavily pretreated patients with CD19+ relapsed or refractory (r/r) chronic lymphocytic leukemia (CLL) or NHL were enrolled in the clinical trial; they received UCB-derived CAR-NK cells with a novel construct incorporating the IL-15 gene and an inducible caspase-9-based suicide gene after lymphodepleting chemotherapy. One of three doses (1×10⁵, 1×10⁶, or 1×10⁷ CAR-NK cells per kilogram of body weight) of CAR-NK cells was infused into different patients; a maximum tolerated dose was not reached. Responses were evident within 30 days: eight of eleven patients had a response, seven (four with lymphoma and three with CLL) of whom exhibited complete remission. With the inclusion of the IL-15 construct, the infused CAR-NK cells persisted at least 12 months in vivo, as measured by a quantitative real-time polymerase-chain-reaction assay. After the infusion of CAR-NK cells, all the patients suffered transient and reversible hematologic toxic events, such as neutropenia, lymphopenia, lymphopenia and anemia, which were mainly caused by the lymphodepleting chemotherapy. However, there were no cases of CRS, neurotoxicity, hemophagocytic lympho-histiocytosis, tumor lysis syndrome or grade 3 or 4 nonhematologic toxicity. Despite the HLA mismatch between patients and their CAR-NK products, no symptoms of GVHD were observed. As a result, the caspase 9 safety switch was not activated because of the absence of severe adverse events in the clinical trial. FT596, a universal off-the-shelf CAR-iNK cell product that contains CD19, hnCD16, and an IL-15 receptor, is currently under investigation in two clinical trials (NCT04555811 and NCT04245722). The aim of the Phase I multi-center study (NCT04555811) is to evaluate the safety of FT596 when given with rituximab as relapse prevention in patients who have undergone auto-HSCT for B-cell lymphoma. In the past 6 months, two new clinical trials (NCT04796675 and NCT04639739) have been registered concerning CD19 CAR-NK cells, indicating that the promising clinical trial results have aroused the interest of researchers in the field of CAR-NK cell immunotherapy and bringing new hope for patients with B-cell malignancies. In addition to CD19 CAR-NK cells, the anti-tumor activation and safety of CD22 CAR-NK cells for B-cell malignancies are also under investigation in clinical trials (NCT03692767 and NCT03824964).

MM, comprising approximately 10% of hematological malignancies, is a clonal plasma cell disorder that produces excess monoclonal immunoglobulin. The disease often features hypercalcemia, renal failure, anemia, and bone lesions (82).

B-cell maturation antigen (BCMA) is widely expressed on plasma cells, myeloma cell lines, and primary myeloma while absent in primary human CD34+ hematopoietic cells (83). CAR-T cells that target BCMA have been mostly used in MM; CD38, CD138, and CD319 (CS1) are also common target antigens for MM (84). Additionally, the NKG2D ligand (overexpressed on MM cells) is a potential target for MM. Because CAR-T-cell therapy has generated considerable enthusiasm in patients with MM, CAR-NK cell therapy for MM is also a focus of research.

FT576 NK cells, a CAR-NK cell type derived from iNK cells, exhibit uniform expression of CD16, anti-BCMA CAR, and IL15-receptor α fusion protein (IL-15RF) and did not express CD38 (62). Preclinical trials proved FT576 NK cells enhanced cytotoxicity and persistence, avoidance of self-fratricide, and prevention of antigen loss when combined with other mAbs (e.g., anti-CD38 mAb) in the treatment of MM (61, 62). Chu et al. indicated that CS-1 CAR-NK92 cells displayed CS1-dependent recognition and killing capacity for both MM cell lines and primary MM cells; this cell therapy prolonged survival in xenograft model mice by inhibiting the growth of MM cells (85). Moreover, NK cells carrying a CD138 CAR exhibited enhanced cytotoxicity against CD138+ human MM cell lines and primary MM cells with increased secretion of granzyme B and IFN‐γ, as well as increased expression of CD107a, both in vitro and in xenograft models (86).

A clinical trial (NCT03940833), conducted by Wuxi People’s Hospital, plans to recruit 20 patients with MM. The purpose of that study is to assess the safety and feasibility of BMCA CAR-NK92 cells for patients with r/r MM.

T-cell malignancies mainly include T-lymphoblastic leukemia and T-lymphoblastic lymphoma, and T-lymphoblastic lymphoma can be further divided into two categories: cutaneous T-cell lymphoma (CTCL) and peripheral T-cell lymphoma (PTCL). Although CAR-T cell therapy has been successfully used in B-cell malignancies, it generally has not shown good results in the treatment of T-cell malignancies because of the shared expression of target antigens between normal and malignant T cells, which may lead to fratricide and immunodeficiency (87). Thus far, molecules used as the targets for CAR-based cellular therapy in T-cell malignancies are CD3, CD4, CD5, CD7, CD30 and CD1a (88–93). Recently, preclinical and clinical trials have been initiated using CAR-NK cells as a therapy for T-cell lymphoma/leukemia.

Chen et al. have generated CD3 CAR-NK92 cells and demonstrated specific and effective lysis of CD3+ human PTCL primary samples, as well as T-cell leukemia cell lines. Moreover, CD3 CAR-NK92 cells controlled the growth of T-ALL in xenogeneic mouse models, with an 87% tumor burden reduction on day 13 and maintenance at this level on day 23 (94). Furthermore, the same group has confirmed the safety and effectiveness of CD4 CAR-NK92 cells against various CD4+ T-cell lymphoma/leukemia cell lines and patient samples (95). Wang et al. compared two CD5 CAR-NK cells with different costimulatory molecules: 4-1BB and 2B4. Their findings suggested that NK-cell-associated activating receptor 2B4 may further enhance the killing activity of CD5 CAR-NK cells against CD5+ malignancies, compared with T-cell-associated activating receptor 4-1BB (96, 97). The median survival time of xenogeneic mice was longer in the CD5-2B4-CAR-NK92 group than in the CD5-4-1BB-CAR-NK92 group: 45.5 days and 58.5 days, respectively (p < 0.05) (97). You et al. first reported two CD7 CAR-NK92-MI cell constructs using the CD7 nanobody VHH6 sequences; they demonstrated that CD7 CAR-NK92-MI cells exert specific cytotoxicity and an inhibitory effect on primary T-ALL cells in a PDX mouse model (98).

Currently, only one clinical trial (NCT02742727) related to T-cell malignancies is registered; the purpose of this trial is to evaluate the safety and effectiveness of CD7 CAR-NK92 cell immunotherapy in patients with CD7+ r/r lymphoma and leukemia. However, this study has not yet been published. The shared antigens between tumor cells and healthy tissues remain a major challenge for CAR-based cellular immunotherapy in T-cell malignancies. Thus, there is a need for approaches to seek out new target antigens expressed exclusively on tumor cells or identify the appropriate therapeutic windows of CAR-NK cells to minimize T-cell depletion (96).

Myelodysplastic syndromes (MDS) and acute myelogenous leukemia (AML) are two mainly myeloid malignancies, characterized by uncontrolled proliferation of undifferentiated myeloid progenitor cell clones (99). With the progression of disease, there is a risk of MDS conversion to AML. Application of CAR-T cell therapy in AML is difficult because nearly all currently known target antigens are expressed both on leukemic blasts and healthy progenitor cells. Therefore, methods using CAR-T cells as salvage therapy or as a bridge to allo-HSCT for AML may circumvent these challenges (100). Potential target antigens (e.g., CD33, Lewis Y, CD123, CD135, CLL1, CD44v6, FRβ, CD38, and CD7) for AML have been summarized by Hofmann et al. (101). Klöß et al. generated CD123 CAR-NK cells and showed excellent capacity for eliminating resistant AML blasts and AML cell lines ex vivo (102). In addition to expression in T-cell malignancies, CD4 antigen is also expressed in a subtype of AML; therefore, CD4 CAR-NK cells also exhibit potent antileukemic activity against CD4+ AML (103). A first-in-man small clinical trial (NCT02944162) of CD33 CAR-NK92 cells for patients with r/r AML has been completed with published data (74). In this clinical trial, each of three patients received doses of up to 5 × 10⁹ CD33 CAR-NK92 cells. Among them, two patients suffered a moderate fever and one patient suffered a high fever after CD33 CAR-NK92 cell infusion, and all of them returned to normal within in two days. Grade I CRS was observed in patient 1 and 3. No other significant adverse effects were observed. Though no conclusion regarding clinical efficacy could be drawn, the safety of using CD33 CAR-NK92 cells for r/r AML patients with high tumor burden was clear.

In addition to the target antigens mentioned above, NKG2D ligands are promising target antigens because of their increased expression in AML and MM, but absence in healthy tissues (104). The safety and capacity of manufactured CAR-T cells that target NKG2D ligands for AML patients and MM patients have been verified in a phase I clinical trial (105). Recently, one new clinical trial (NCT04623944) was registered to determine the safety and tolerability of an experimental therapy known as NKX101 (allogeneic CAR-NK cells targeting NKG2D ligands) in patients with r/r AML, r/r MDS or intermediate (high/very high)-risk MDS.

Currently, similar to CAR-T cell therapy, clinical trials about CAR-NK cell therapy are mainly focused on r/r patients only. Several studies have suggested that the infusion of unmodified NK cells is feasible to maintain complete remission (CR) status for MDS or AML patients (55, 106, 107). Besides, adoptive NK cell transfer either alone or combined with chemotherapy could also be used as a post-CR consolidation strategy in AML (108, 109). A phase II trial that transfusing NK cells as a consolidation therapy in pediatric patients with AML has been conducted, and the result is still awaited (110). Additionally, one study also suggested that infusing NK cells could also consolidate incomplete engraftment in patients after haploidentical HSCT (19). Therefore, it is reasonable to assume that in addition to r/r patients, CAR-NK cells may also hold the potential to be used in consolidation and maintenance stages for treating myeloid malignancies. We believe that the potential clinical applications of CAR-NK cell therapy will be further extended in the future.